What's New — ClearSpeak App Updates & Changelog

✨ New Features

🥊 NewDebate Arena — A new premium practice tool that trains the one thing no other module touched: argumentation under pressure. You're handed a position to defend, record a claim-first response on a countdown, then face a rebuttal and counter it. Mico scores each round on claim clarity, evidence quality, and delivery conviction. Ideal for debates, stakeholder pushback, interview follow-ups, and high-pressure Q&A.
🏅 New8 Debate achievements — Unlock milestones as you refine your argumentation: First Claim, Clean Rebuttal, Iron Conviction, and more. All premium-only and tracked in your profile.
💬 New/debate slash command in Mico chat — Launch a Debate Arena round on any topic directly from chat. Great for "quiz me on X before my 4pm" style reps.
🗣️ NewNeural voice across onboarding & cues — Edge-TTS is now the primary voice provider everywhere (onboarding, lesson phase transitions, Mico chat), with graceful fallback to device voice when offline. A new "Neural Voice" indicator lives in Settings.

🔧 Improvements

🎧 ImprovedCoach session audio stability — Live coaching cues now correctly use the device voice during active recording to prevent iOS audio session conflicts. Recording no longer cuts out when Mico speaks mid-session.
📚 ImprovedLesson screen narration — Every Learn phase now has a "Read aloud" option and Mico narrates phase transitions automatically using the neural voice.
🧭 ImprovedMico chat memory — Chat context assembler now pulls richer recent-session signals (pace trend, dominant filler, latest freeze moments) so Mico answers with concrete data instead of generic tips.

🐛 Bug Fixes

🔧 FixedFixed sticky CTA bar covering the last row of the footer on mobile devices.
🔧 FixedFixed anchor links (FAQ, Download, How It Works) scrolling behind the fixed header.
🔧 FixedTightened Debate Arena and Coach Session audio handoff so STT and playback never overlap.

✨ New Features

🧠 NewAI Coach Memory — The coach now builds a private profile across every session: your dominant filler word, when your pace drifts, whether your endings fade, restart loop patterns, and more. This profile is rebuilt after each session and used to personalize your daily plan. Stored 100% on-device.
🎓 NewSkill Grade System (A–F) — Pace, fillers, intonation, and structure now each have a live letter grade that updates based on your sessions. Complete a filler session with a rate above your goal? Grade drops. Nail a coach session with strong endings? Grade rises. Your daily plan automatically targets your lowest grade.
📍 NewFiller Hotspot Detection — The filler engine now identifies where your fillers cluster: sentence openings, topic transitions, or high-pressure moments. Your daily drill recommendation is routed to the specific drill that targets that hotspot — "Clean Open Drill," "Smooth Transition Drill," or "Pressure-Proof Drill."
🗓 NewAdaptive "Today in 3 Steps" — The daily practice plan is now rebuilt from your current skill grades and coach memory after every session. If your structure grade drops (quiz failure), a targeted drill appears. If your dominant filler word emerges, the insight surfaces in your context message.

🔧 Improvements

🎯 ImprovedPersonalized thresholds — WPM pace targets and filler rate goals are now calibrated to your personal baseline and stated goals (set during onboarding), not fixed industry averages. A user targeting 100–130 WPM gets a different readiness assessment than one targeting 140–165.
📝 ImprovedQuiz results now feed skill grades — Scoring below 60% on a quiz now lowers your structure skill grade, making the weak area visible in your daily plan and surfacing a targeted review drill. Previously, quiz performance was tracked but never acted on.
🔄 ImprovedCoach preferences persist between sessions — Your last-used coach mode (gentle / balanced / direct) and cue modality (voice / text / adaptive) are now saved automatically after each session. No more resetting to defaults every time.
📅 ImprovedRecency-weighted weak area detection — Weak area analysis now prioritizes your last 14 days of sessions. Patterns from 4 weeks ago no longer override recent improvement. If you've cleaned up your pace recently, the app now notices.

✨ New Features

🗓 New30-Day Journey Timeline — A visual day-by-day map of your progress across any program. See where you are, what's coming, and how far you've come. Added to onboarding and program detail screens.
🗂 NewModule filter tabs in the practice library — Filter all 18 modules by category: Fluency, Confidence, Clarity, Accent, Interview, and more. Makes it much faster to find the right session for your mood.
📊 NewReading progress indicator on session screens — A subtle bar shows how far through a session you are. Helps with pacing during longer Table Topics or script read-aloud sessions.
🎙 NewTap-to-count filler fallback — If speech recognition is unavailable or the user prefers manual tracking, they can tap the screen each time they notice a filler. Counts are merged with auto-detected results for the session summary.

🔧 Improvements

⚡ ImprovedOn-device AI analysis latency reduced by ~40ms on average on recent iOS and Android devices. Real-time filler nudges now feel more immediate.
🗣 ImprovedFiller detection engine expanded with 12 new regional variants (e.g. "innit", "you get me", "know what I mean", "sorta", "kinda like"). More accurate for speakers outside North America.
🎨 ImprovedSession summary screen redesigned — cleaner data hierarchy, filler heatmap by sentence, and a one-tap "Replay your worst moment" button that jumps to the highest-filler segment of your session.

🐛 Bug Fixes

🔇 FixediOS audio session conflict — expo-av recorder and speech recognition could not run simultaneously, causing silent sessions on certain iOS 17 devices. The app now uses one audio path at a time, with a reliable handoff sequence. No more silent recordings.
🔢 FixedDouble-counting bug in filler detection — Certain multi-word fillers ("you know", "I mean") were counting each word separately AND as a phrase, inflating filler counts. Detection engine now uses a deduplicated match set.
💬 FixedApostrophe normalization — words like "it's" and "don't" were not being tokenized correctly by the filler engine, causing some contractions to be missed. Apostrophes are now stripped before tokenization.

✨ New Features

📧 NewEmail updates opt-in — Users can now subscribe to ClearSpeak updates, new program announcements, and tips directly from the landing page or settings screen.
🌐 NewLanding page launched at myclearspeak.app — full redesign with 18 sections, interactive module library, persona selector, and comparison table. Built with on-device privacy messaging front and center.
🎯 NewHero result metrics displayed in onboarding — users now see "73% avg filler drop" and "Day 7 results" before starting, setting a concrete, motivating target.
🔔 NewDaily reminder notifications — Opt-in push notifications to maintain streaks. Notification time is fully customizable. Respects Do Not Disturb and Focus modes.

🔧 Improvements

🏃 ImprovedCold start time reduced by ~30% on mid-range Android devices. The on-device model now loads incrementally rather than blocking the main thread at launch.
🌙 ImprovedXP system rebalanced — earning XP now feels more meaningful. Streaks above 7 days give bonus XP. Long sessions (over 8 minutes) receive a consistency multiplier.
♿ ImprovedAccessibility overhaul — all interactive elements now have proper focus states and ARIA labels. VoiceOver (iOS) and TalkBack (Android) tested across all primary screens.

🐛 Bug Fixes

🔄 FixedSTT (speech-to-text) permission request was being triggered before checking availability, causing a crash on first launch on certain Android 12 devices.
⏱ FixedPost-recording delay too short (250ms) was causing the final transcript to be cut off before the filler analysis completed. Increased to 800ms with a visual loading indicator.
📱 FixedStreak counter not resetting at midnight in UTC+12 timezones. Streak calculations now use device local time consistently.

ClearSpeak launches on iOS and Android. Built from the ground up for one purpose: daily, private, deliberate speech practice — 5 minutes at a time.

🚀 Launch Features

🎓 17 guided learning programs — ClearSpeak Core (7 days), Confidence Builder, Presentation Power, Fluency Mastery, Public Speaking Mastery (90 days), English Confidence Journey (168 days), and more. Each program is structured with clear day-by-day progression.
🗂 18 practice modules — Filler Word Coach, Table Topics (100+ prompts), Shadowing (50+ clips), Tongue Twisters (160+ exercises), Debate Arena, Warm-Up Studio, CBT Tools, Breathing Sanctuary, and more.
🤖 Real-time on-device AI — 150+ filler word terms detected in real time. Pacing analysis, ending strength detection. All processing happens locally. Works in Airplane Mode. No voice data ever sent to any server.
🔬 Speech Lab baseline — 2-minute free-speech baseline at onboarding creates a personalized starting snapshot. Used to recommend the right programs and track improvement over time.
🏆 Streak & XP system — 7 progression levels from Bronze to Diamond. Daily streaks, weekly challenges, and XP rewards keep practice consistent. Proven behavior-design principles (implementation intentions, variable reward) built into the loop.
🌍 English Confidence Journey — 168-day program for non-native English speakers. 4 progressive stages: Foundation, Everyday, Social & Professional, Advanced Expression. Accent training, phoneme drills, shadowing with native speakers.

What's New in ClearSpeak